Reading Notes: Coding-Agent Harnesses, Evals, and RL
Published:
A few recent papers I have been reading on agent process discipline, workflow migration, multi-agent code generation, and RL for code.
Published:
A few recent papers I have been reading on agent process discipline, workflow migration, multi-agent code generation, and RL for code.
Published:
Why we moved Kiro’s code editing from string matching to AST-based structural operations — and what it bought us in tokens, latency, and reliability.